In recent years, I've carried a 'compact' camera with me on major rides, to record the sights. This year, however, when riding in the Alps, I'm planning on relying on my mobile. Recent advances in mobile phone cameras include some fairly sophisticated High Dynamic Range(HDR) capabilities. Multiple exposures are taken automatically, and merged to maximise the perceived dynamic range in the resulting picture. This is just what you need to capture photos with a large contrast range. And that includes landscapes with bright skies.
 |
| View from the top of Dean Hill, direct from my phone |
First experiments, from a group ride last Sunday, look quite promising. As a result, I'm planning on leaving my camera behind on big climbs, and using the extra space for more food!
Having a bit of history in image processing, I was interested in how the newer HDR approaches work. Conventional HDR involves taking a series of individual images while bracketing the exposure. Some shots will be under exposed while others will be over exposed. The combination of the data from the set of images, using the appropriate maths, leads to an image that better represents the full range of colour and brightness. Images in the sequence need to be registered with one another, of course, since there may be slight movements of the camera between individual frames. But that's the same task as stabilisation in video recording, and that has been available in real time for quite a while. The disadvantage with conventional HDR is the time taken to capture the multiple frames. This manifests itself as shutter lag to the photographer.
The newest approach to HDR appeared in mass market phones with Google's Pixel range. In Android, this is now known as HDR + Auto. As soon as the camera is running, it is actually capturing a continuous stream of frames, as if it were in video mode. These are stored in a buffer. Every frame is deliberately under exposed, so that they can be combined without 'blowing out' bright areas. When the shutter is pressed, the camera records the time, and retrieves the appropriate set of frames from the buffer. Because the data is being captured continually, there is no shutter lag at all. The challenge, of course, is that because no frames are over exposed, the data in the shadow areas is always noisy. It turns out, however, that better results can be obtained in the shadows by combining a time sequence of frames, than by boosting the exposure. In particular, colours can actually be captured more accurately in shadows. The whole process takes a lot of computation, of course, and has only become feasible with the latest processors. HDR + is an example of the new trend towards computational photography and software defined cameras, which use the basic data from the sensor in new ways. In many respects, its closer to what our own visual systems do, in post processing the data from our eyes, which are not the best visual sensors, to generate detailed representations of the world.
There are likely to be glitches, from time to time, but it definitely seems to be worth trying in anger.
Having a bit of history in image processing, I was interested in how the newer HDR approaches work. Conventional HDR involves taking a series of individual images while bracketing the exposure. Some shots will be under exposed while others will be over exposed. The combination of the data from the set of images, using the appropriate maths, leads to an image that better represents the full range of colour and brightness. Images in the sequence need to be registered with one another, of course, since there may be slight movements of the camera between individual frames. But that's the same task as stabilisation in video recording, and that has been available in real time for quite a while. The disadvantage with conventional HDR is the time taken to capture the multiple frames. This manifests itself as shutter lag to the photographer.
The newest approach to HDR appeared in mass market phones with Google's Pixel range. In Android, this is now known as HDR + Auto. As soon as the camera is running, it is actually capturing a continuous stream of frames, as if it were in video mode. These are stored in a buffer. Every frame is deliberately under exposed, so that they can be combined without 'blowing out' bright areas. When the shutter is pressed, the camera records the time, and retrieves the appropriate set of frames from the buffer. Because the data is being captured continually, there is no shutter lag at all. The challenge, of course, is that because no frames are over exposed, the data in the shadow areas is always noisy. It turns out, however, that better results can be obtained in the shadows by combining a time sequence of frames, than by boosting the exposure. In particular, colours can actually be captured more accurately in shadows. The whole process takes a lot of computation, of course, and has only become feasible with the latest processors. HDR + is an example of the new trend towards computational photography and software defined cameras, which use the basic data from the sensor in new ways. In many respects, its closer to what our own visual systems do, in post processing the data from our eyes, which are not the best visual sensors, to generate detailed representations of the world.
There are likely to be glitches, from time to time, but it definitely seems to be worth trying in anger.
No comments:
Post a Comment